Crowdfunding has been recently rising as financing channel and showed rapid growth by integrating with social media. As of 2018, global crowdfunding market size was estimated as $ 9.37 billion, and Korea crowdfunding market size was about $ 110 million. However, the probability of crowdfunding failure showed more than 38%, which gives huge burden for participants (i.e., makers, investors, platforms). So, to prevent the failure and protect participants from their loss of time and money, predicting the success of the funding in the early step is crucial. Therefore, this study aims to build a model to predict whether the crowdfunding project will success or fail. Compare to the previous studies that they used data after the end of crowdfunding, we collected data seven days before the project ends. We used data from crowdfunding site ‘Wadiz’, by collecting comment data and funding information as predict variable. Then we applied machine learning methods such as Decision Tree, Support Vector Machine, Naive Bayes, AdaBoost, Gradient Boosting, Random Forest, and MLP. As a result, Gradient Boosting showed more than 90% accuracy, and Support Vector Machine showed the highest precision score (0.95). Also, this study has a practical implication of predicting funding success in the early stage of crowdfunding by developing a prediction model based on machine learning.